System and algorithm design for a new generation tree-ring image analysis system

We introduce a new generation of a semiautomated tree-ring image analysis system and new image analysis algorithms. System design, new tree-ring image registration, ring boundary identification, and average ring width and brightness profile measurement algorithms are described. By using a pyramid multiscale image registration algorithm, the program gains registration speed with little sacrifice of accuracy. For boundary identification, two new assumptions and associated nonring feature suppression, edge tracking, and fragment linking methods are proposed, such that most of the complicated ring boundaries can be identified successfully. To obtain more detailed information about the ring growth changes with time, an average brightness profile measurement algorithm is implemented. Additionally, automatic geometry warping is introduced to the average ring width and brightness profile measurement to handle abnormal ring cases. To allow manual override of the results, a new user intervention design is also implemented and described.